Install zfp with Homebrew, apt, dnf, MacPorts, Nix, pacman, zypper

Project links

• Homepage
• Repository
• Upstream docs
• Package manager page

Commands and aliases

• zfp

Project history

zfp was originally developed by Peter Lindstrom at Lawrence Livermore National Laboratory. Its foundational publication is Lindstroms 2014 IEEE TVCG paper on fixed-rate compressed floating-point arrays, which addressed a key limitation of variable-length floating-point compression: difficult memory management and poor random access.

The software grew from an algorithmic compressor into a broader compressed-array package. The documentation describes four main components: the zfp number format for fixed-size blocks, C++ compressed-array classes, a C library for streaming compression of whole or partial arrays, and a command-line executable for binary numeric array files.

Modern zfp keeps both lossy and lossless use cases in view. The LLNL project page describes high-throughput read/write random access, optional error-bounded compression, CPU and GPU execution back ends, and bindings for Python, Fortran, and other environments.

Adoption history

zfp adoption is tied to scientific data reduction. LLNL and Exascale Computing Project material describe zfp as part of the ALPINE/zfp effort, aimed at reducing I/O and storage pressure in exascale simulations by compressing floating-point arrays while preserving user-controlled error bounds.

The library also spread through integration points. zfp documentation lists application support in ADIOS, BLOSC, HDF5 through H5Z-ZFP, Intel IPP, MVAPICH2-GDR, OpenInventor, and OpenZGY; the H5Z-ZFP docs describe a registered HDF5 filter using zfp for bitrate, accuracy, and precision targets.

How it is used

Users choose compression modes according to workflow constraints. Fixed-rate mode gives predictable size and block random access; fixed-precision and fixed-accuracy modes trade fixed size for precision or absolute-error control; reversible mode supports bit-for-bit reconstruction for supported data and execution paths.

In application code, zfp arrays can act as compressed replacements for native C/C++ arrays or std::vector-like storage when random element access matters. For file or stream workflows, the C API and CLI compress integer or floating-point array data, while Python and Fortran bindings make the same compressor reachable from scientific scripting and simulation environments.

Why package nerds care

zfp is significant because it packages a research-grade numerical representation as normal developer infrastructure: a C/C++ library, CLI, language bindings, HDF5 plugin ecosystem, and HPC distribution target all orbit the same compressed-array format.

For package catalogs, it sits at the intersection of compression, numerical correctness, and hardware-aware performance. Unlike general compressors, zfp exposes scientific knobs such as rate, precision, and accuracy, so the installed package is a policy surface for data fidelity as much as a space-saving tool.

Timeline

• 2014: Peter Lindstrom publishes the fixed-rate compressed floating-point arrays paper that underpins zfp.
• 2016: H5Z-ZFP documentation appears with LLNL-CODE-707197, reflecting zfp use as an HDF5 compression filter.
• 2022: LLNL announces zfp 1.0.0, describing the package as a compressed numerical array library with C/C++, Python, and Fortran surfaces.
• 2020s: Exascale Computing Project material places zfp in the ALPINE/zfp data reduction and visualization workflow for large simulations.

Related projects

• H5Z-ZFP is the HDF5 compression filter built on zfp.
• zfPy is the Python interface to zfp.
• ALPINE is the ECP visualization and analysis effort paired with zfp for exascale data reduction.
• SZ and MGARD are neighboring scientific lossy-compression projects often discussed alongside zfp.

Sources

• https://computing.llnl.gov/projects/zfp
• https://github.com/LLNL/zfp
• https://h5z-zfp.readthedocs.io/en/latest/
• https://software.llnl.gov/news/2022/08/01/zfp-1.0.0/
• https://www.exascaleproject.org/highlight/alpine-zfp-addresses-analysis-visualization-and-data-reduction-needs-for-exascale-science-applications/
• https://www.osti.gov/biblio/1228014
• https://zfp.readthedocs.io/en/release1.0.0/modes.html
• https://zfp.readthedocs.io/en/release1.0.1/introduction.html